The New Quantum Era on Mythryl: “Individual layers of Meta's Llama 3.1 8B are running on a 156-qubit…”

Individual layers of Meta's Llama 3.1 8B are running on a 156-qubit IBM quantum processor right now — not a simulation, an actual inference step. Román Orús, the physicist who co-invented the tensor network math making this possible, explains what it proves and what it doesn't.

newquantumera.comQuantum-Inspired AI and Tensor Network Compression with Román OrúsTensor networks were invented to study quantum entanglement in condensed matter physics — so how did they end up compressing billion-parameter language models and running inside a live LLM on an IBM quantum processor?

7/9/26, 9:06 pm2 reposts3 likes

Post